> A perhaps superfluous remark, but is the diskAfter verification, the drive where the jobs are running isn't full, we have 300GB of free space left.
> Most issues can be found in the shadowlog on the pool manager, can you find the job in the ShadowLog and send it on?
I suppose you want to say ShadowLog on submission machine? Sorry but the rotation of log has already delete the part concerning
What I found to force jobs to run is to restart condor (/etc/init.d/condor restart) on the submission machine. Each time I do that
a number of jobs equivalent of number of CPU in the pool begin to run. But It's not a long terme solution for me ;-)